Article pubs.acs.org/JACS
Perfluoroalkyl Cobalt(III) Fluoride and Bis(perfluoroalkyl) Complexes: Catalytic Fluorination and Selective Difluorocarbene Formation Matthew C. Leclerc,† Julia M. Bayne,† Graham M. Lee,† Serge I. Gorelsky,† Monica Vasiliu,‡ Ilia Korobkov,† Daniel J. Harrison,† David A. Dixon,‡ and R. Tom Baker*,† †
Department of Chemistry and Biomolecular Sciences and Centre for Catalysis Research and Innovation, University of Ottawa, 30 Marie Curie, Ottawa, Ontario K1N 6N5, Canada ‡ Department of Chemistry, The University of Alabama, Tuscaloosa, Alabama 35487, United States S Supporting Information *
ABSTRACT: Four perfluoroalkyl cobalt(III) fluoride complexes have been synthesized and characterized by elemental analysis, multinuclear NMR spectroscopy, X-ray crystallography, and powder X-ray diffraction. The remarkable cobalt fluoride 19F NMR chemical shifts (−716 to −759 ppm) were studied computationally, and the contributing paramagnetic and diamagnetic factors were extracted. Additionally, the complexes were shown to be active in the catalytic fluorination of p-toluoyl chloride. Furthermore, two examples of cobalt(III) bis(perfluoroalkyl)complexes were synthesized and their reactivity studied. Interestingly, abstraction of a fluoride ion from these complexes led to selective formation of cobalt difluorocarbene complexes derived from the trifluoromethyl ligand. These electrophilic difluorocarbenes were shown to undergo insertion into the remaining perfluoroalkyl fragment, demonstrating the elongation of a perfluoroalkyl chain arising from a difluorocarbene insertion on a cobalt metal center. The reactions of both the fluoride and bis(perfluoroalkyl) complexes provide insight into the potential catalytic applications of these model systems to form small fluorinated molecules as well as fluoropolymers.
■
INTRODUCTION Transition metal complexes bearing fluoride or fluorocarbon ligands have attracted considerable interest because they are used to mediate/catalyze C−F or C−RF bond-forming reactions, which are highly important in the pharmaceutical, agrochemical, and advanced materials industries.1,2 Despite this widespread interest, the fundamental chemistry of these species is considerably less developed than that of analogous hydrocarbon compounds. In particular, reports of complexes bearing two fluorinated ligands (i.e., one perfluoroalkyl and one fluoride, or two perfluoroalkyls) are very rare, with most examples belonging to second or third row metals.3 Recently, examples of Ni complexes bearing two perfluoroalkyl ligands have been reported.4 There are synthetic challenges associated with preparing such complexes: The most direct approach would be via oxidative addition of the C−F or C−C bond of a perfluoroalkane (CF4, C2F6, C3F8, etc.) to a low-valent metal, but the inert nature of perfluoroalkanes makes this route inaccessible.1 Here, we use alternative synthetic routes to access the products of the hypothetical oxidative addition reaction between perfluoroalkanes and first row metals. Our general strategy is to utilize the oxidative addition of iodoperfluoroalkanes (RF−I) to install the first perfluoroalkyl group on the metal, followed by exchange of the iodide ligand for either a fluoride or a trifluoromethyl group (Scheme 1). Oxidative © XXXX American Chemical Society
Scheme 1. Alternative Synthetic Route to Transition Metal Fluorides and Perfluoroalkyls
addition of RF−I to metal complexes has been shown to proceed for group 9 metals,5 and methods for converting [M]− X (X = halide) to [M]−F6 or to[M]−CF32,7 are known. Reactions between the inexpensive and commercially available cobalt(I) complex CpCo(CO)2 (Cp = η5-cyclopentadienyl) and RF−I (RF = CF3 and CF2CF3) furnish cobalt(III) Received: August 31, 2015
A
DOI: 10.1021/jacs.5b12003 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX
Article
Journal of the American Chemical Society complexes CpCo(RF)(I)(CO).8 Substitution of the carbonyl ligand with a phosphine is facile and leads to the series of isolable starting materials CpCo(RF)(I)(L) (1−4), as shown in Scheme 2.9
formed from the reaction of the one-electron reductant cobaltocene with an excess of perfluorodecalin in toluene at low temperature. CoF3 is commercially available, although it is often too reactive to promote transformations in a selective manner. Of these three systems, only cobaltocenium fluoride and the cyclometalated cobalt fluoride are truly organometallic complexes, but they do not offer any opportunity for varying the ligand environment on cobalt because their scaffolds are limited as a result of the conditions of forming the fluorides, contrary to complexes 5−8, which offer the ability to modify both the nature of the phosphine ligands and the perfluoroalkyl ligands on cobalt. X-ray structural studies confirm that complexes 5−8 are welldefined monomeric CoIII fluorides featuring cyclopentadienyl, phosphine, and perfluoroalkyl ligands (Figure 1). The Co−F
Scheme 2. Synthetic Scheme for Phosphine Substitutions
In recent reports, we described the two-electron reduction of complexes 1−4 with sodium to furnish a series of nucleophilic CoI perfluorocarbene complexes, and demonstrated [2 + 2] cycloaddition reactions with tetrafluoroethylene.10 The resulting cobalt(III) perfluorometallacyclobutane complexes reacted with both Lewis and Brønsted acids to give ring-opening/ isomerization products. However, the chemistry of cobalt(III) systems with multiple perfluorinated ligands remains largely unexplored, and herein we expand that area.
■
RESULTS AND DISCUSSION Synthesis and Characterization of Perfluoroalkyl Cobalt Fluorides. Seeking to isolate the hypothetical products that would arise from the oxidative addition of perfluoroalkanes to a cobalt center, we opted for a pathway involving the substitution of iodide for fluoride, using a method previously reported by Hughes et al. to afford analogous perfluoroalkyl IrIII fluorides.6b Reactions of complexes 1−4 with 3 equiv of AgF in dichloromethane at room temperature over 20 h in the absence of light afforded perfluoroalkyl CoIII fluoride complexes of the general formula CpCo(RF)(F)(L) (Cp = η5-cyclopentadienyl, RF = CF3 and CF2CF3, L = PPh3 and PPh2Me) (5−8) in 68− 91% isolated yield as dark-green solids (Scheme 3). Complexes
Figure 1. Crystallographic representations of 5 (top left), 6 (top right), 7 (bottom left), and 8 (bottom right) with 30% probability thermal ellipsoids. Hydrogen atoms are omitted for clarity. One molecule of acetonitrile has been removed from 5. Sample of 6 crystallized with two molecules in the unit cell. Selected bond lengths and angles are presented in Table S1.
Scheme 3. Synthesis Scheme for Cobalt(III) Fluorides
bond distances in complexes 5−8 range from 1.86 to 1.88 Å (Table S1), similar to the value of 1.89 Å found in CoF3.19 For perfluoroethyl complexes 7 and 8, the Cα−F bond distances (avg. 1.378(2) and 1.393(2) Å) are significantly longer than Cβ−F (avg. 1.326(2) and 1.333(2) Å) as observed previously for an Ir analog.6b The Co−P distances are approximately 0.04 Å shorter with PPh2Me as compared to PPh3 because the former is known to be a slightly more basic donor ligand. Moreover, the Co−C bond distances are shorter for the trifluoromethyl ligand versus the perfluoroethyl fragment by 0.2 Å for the PPh3 derivatives and 0.4 Å for the PPh2Me examples. Seminal work by Stone et al. has established that [M]−C bonds are shorter with perfluoroalkyls than with analogous hydrocarbons, an effect observed in this system as well.20 Recently, another example of a transition metal simultaneously bearing a fluoride and a perfluoroalkyl was reported that features a bis(trifluoromethyl) nickel dimer with bridging fluoride ligands.21 DFT calculations were used to gain insight into the electronic structure of 5 as a representative example. TDDFT calculations at the B3LYP/TZVP level with the SMD solvent model22 reproduced the electronic absorption spectrum in CH2Cl2 well, with two principal experimental bands at 16
5−8 were characterized spectroscopically and structurally, and the results were further analyzed by density functional theory (DFT)11 calculations with the B3LYP12,13 and PW9114,15 exchange-correlation functionals and polarized double- and triple-ζ basis sets. Structurally, complexes 5 and 6 represent the expected products arising from the oxidative addition of perfluoromethane to cobalt, whereas complexes 7 and 8 are those that would arise from the same type of reaction with perfluoroethane. As previously mentioned, these oxidative addition reactions are not feasible; thus, it is necessary to utilize other synthetic methods to obtain such complexes. Cobalt fluorides are uncommon in the literature, and the few that have been presented mostly feature cobalt in either the +1 or the +2 oxidation state.16 There are only three examples featuring cobalt in the +3 oxidation state: cobaltocenium fluoride, CoF3, and an example from Klein et al. with a cyclometalated complex featuring azine as an anchoring group.17 Cobaltocenium fluoride was synthesized by Richmond et al. in 1994,18 and has been applied to several stoichiometric fluorination reactions. This extremely hygroscopic reagent is B
DOI: 10.1021/jacs.5b12003 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX
Article
Journal of the American Chemical Society 300 cm−1 (263 M−1 cm−1, calcd = 15 600 cm−1) and 21 800 cm−1 (1190 M−1 cm−1, calcd = 21 700 cm−1). (See Figure S1 and the band assignments in the Supporting Information.) Relative to typical CoIII octahedral inorganic complexes,23 the high intensities of these two absorption bands indicate significant charge-transfer character in the corresponding electron excitations. Calculated Mayer bond orders24 for 5 provide values for Co−Cp (2.37), Co−PPh3 (0.98), and Co− CF3 bonds (0.91) that are unsurprising. However, the value for the Co−F bond (0.61) indicates significant ionic character in this metal−ligand interaction and that the Co−F is the least covalent among the metal−ligand bonds. The 19F NMR spectra of 5−8 exhibit extreme upfield resonances for the fluoride ligands ranging from δ −716 to −759 ppm. These shifts are significantly upfield from the analogous Ir complexes previously reported by both Hughes et al.6b (δ(19F) = −437 to −446 ppm) and Bergman et al.25 (δ(19F) = −413 to −415 ppm). To the best of our knowledge, these represent the most upfield resonances reported for a 19F NMR signal. The resonances at half-height are very broad (900−1900 Hz) and featureless, presumably because of the fluorides being bound to 59Co, a nuclide with a spin of 7/2, a natural abundance of 100%, and a large quadrupolar coupling constant of 42.0 × 10−30 m2, all of which contribute to a significant broadening of the fluoride signal. The addition of molecular sieves to an NMR sample of 5−8 did not affect the broadness of the fluoride signals, indicating that the signal is not broadened artificially by the presence of moisture. From the results of DFT computational studies, we are now able to understand the unique nature of these chemical shifts. The results for all of the calculated Co−F chemical shifts and their diamagnetic and paramagnetic tensor components are shown in the Supporting Information. There are minor quantitative differences between the three sets of chemical calculations but not qualitative differences. There is reasonable agreement with experiment for the CF3 and CF2 chemical shifts with differences of up to 30 ppm, which is typical of such fluorine NMR calculations. The differences between the experimental and the calculated shifts for the F bonded to the Co are larger by 30−100 ppm depending on the method, with the BLYP/TZVP2 results being the closest to experiment for this shift. The magnitudes of the calculated shifts for the Co−F were found to be very sensitive to the bond distance, suggesting why the difference between the calculated and experimental values for this shift can be large. For L = PPh3 and R = CF3, the calculations predict a small value for the 19F shift of the CF3 group (ca. −20 ppm as compared to the experimental value of −2 ppm), so the difference in the diamagnetic and paramagnetic components are comparable to those of the standard CFCl3 (BLYP/TZ2P σ(standard) = 118.8 ppm) with the diamagnetic component larger than the paramagnetic component. The 19F chemical shift for the F bonded to the Co is large and negative, resulting from the fact that the diamagnetic and paramagnetic components have the same sign, both shielding. The paramagnetic component is larger than the diamagnetic component. We note that the diamagnetic shielding component for the F bonded to C and of the F bonded to Co are very similar, within ∼10 ppm, so the large changes are due to the differences in the paramagnetic components between the “normal” value for the F in the CF3 group and the value predicted for the F bonded to Co. The fact that the paramagnetic component tensor has the same sign as the diamagnetic component tensor has been noted
previously for ClF because of mixing of the appropriate π orbitals with the σ* orbital in the presence of a magnetic field.26 Although F2 has the same mixing interactions, the presence of symmetry prevents the paramagnetic component from being shielding. The high-lying occupied and low-lying unoccupied molecular orbitals (HOMO and LUMO, respectively) for CpCo(CF3)(F)(PH3) and CpCo(CF3)(F)(PPh3) are shown in the Supporting Information. The orbitals are essentially the same for both compounds. The HOMO, HOMO-1, and HOMO-2 are lone pairs on the F bonded to Co interacting with different d orbitals on the Co. For the Co contribution, the HOMO is the dx2y2, the HOMO-1 is the dz2, and the HOMO-2 is the dxy. The LUMO is the Co−F σ* orbital with the dxz on the Co, and the LUMO+1 is predominantly the Co−C σ*. Thus, the HOMO, HOMO-1, and HOMO-2 serve as the equivalent to the π-type orbitals in ClF, and the LUMO is the equivalent of the ClF σ*. It is the interaction of these orbitals in the presence of a magnetic field that leads to the paramagnetic component being shielding, similar to what is found for ClF. Reactivity of Fluoride Complexes. The importance of fluorinated organic substrates has been amply demonstrated.27 Efficient, reliable techniques for the introduction of fluorine into such products have been the subject of widespread research for many years.28 Consequently, and encouraged by the ionic character of the Co−F bonds in our system, we sought to determine the ability of these cobalt systems to fluorinate simple organic compounds. Reactions with p-toluoyl chloride were explored as a potential route toward fluorination to form p-toluoyl fluoride. Gray et al. have recently demonstrated this reaction in stoichiometric fashion, proceeding through halide metathesis with cyclometalated iridium fluoride complexes.29 Stoichiometric reactions with complex 6 in C6D6 showed clean and essentially complete conversion of the starting substrate within 2 h and formation of the p-toluoyl fluoride product, proceeding through overall halide metathesis with the cobalt fluoride complex. Prompted by the initial results of these stoichiometric reactions, we aimed to develop a catalytic process whereby, starting with the iodide complex 2, the fluoride complex 6 could be generated in situ by the presence of an excess of AgF. Control experiments convincingly demonstrated that stoichiometric reactions between p-toluoyl chloride and the fluoride sources AgF, CsF, KF, and CoF3 gave minimal conversion of the starting reagent to the target compound overnight in dichloromethane (
